Measuring Serverless Apps in Amazon Bedrock
summary
Shows concrete instrumentation and a working demo: a small serverless app deployed to AWS that logs model invocations, tokens and responses to CloudWatch, traces with X‑Ray, and records audit events with CloudTrail, enabling cost and latency comparison across models called through Amazon Bedrock. Dashboards surface invocation count latency tokens and traces map request timing through Lambda to the model call. Repo deploys, runs, and produces measurable differences so decisions on model choice and prompting are based on numbers not buzzwords.
AI-written summary · about
chapters (8)
- 0:00Introduction and Speaker Introduction
- 1:06Challenges with Generative AI Models
- 2:43Deterministic vs. Probabilistic Systems
- 7:39Key Objectives for AI Applications
- 10:10Building Blocks of Generative AI Applications
- 19:39Implementing Observability in AI Applications
- 22:40Demo and Practical Implementation
- 27:45Conclusion and Closing Remarks
related talks
22mObservability in AI-Powered Applications with Amazon Bedrock
34mObservability in Serverless Application
41mLive Coding on Observability for Serverless
49mMonitoring serverless applications on AWS: the monitoring monster becomes your friend.
39mGenerative AI Solution with AWS Serverless and Bedrock
04mServerless applications, Microservices, AWS
39mObservability in DevOps
47mAdvanced Serverless Architectural Patterns on AWS